Web Case Study
Talk with PDFs
An AI document assistant with a custom RAG pipeline built from scratch — no LangChain — using Vercel AI SDK and pgvector. Streams AI responses directly into a chat UI with per-locument conversation state. Inngest background jobs automate PDF parsing and embedding generation.
Key Features
Custom RAG pipeline with pgvector — no LangChain
Streaming AI responses in real-time chat UI
Per-locument conversation state management
Background PDF parsing & embedding generation via Inngest
Contextual answers using OpenAI + Vercel AI SDK